{"id":"8a71d934-46f2-495c-a156-8530b17ed370","arxiv_id":"2507.20818","paper_version":1,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":4.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":3,"one_line_summary":"Experimental fringe-field mapping of three 3 T MRI scanners shows maximum field differences of 0.5 T and gradient differences of 1.64 T/m between different manufacturers, and 0.17 T and 0.5 T/m within the same manufacturer.","lead":"This study measured the magnetic field in the rooms around three 3 Tesla MRI scanners and built three-dimensional maps from those readings. The maps show that the fields differ between scanners, even for two scanners from the same manufacturer, which matters for staff safety planning.","discovery_kind":"new_application","skeptic_critique":{"model":"deepseek-v4-flash","headline":"Same-manufacturer differences rely on mirrored (unmeasured) left-side data; the paper concedes the symmetry assumption may fail.","rationale":"The reader's weakest_assumption identifies the y-axis mirroring as the key risk, and I agree. The headline same-manufacturer numbers (MRI-2 vs MRI-3) in Tables 3 and 4 occur at negative x coordinates, which are not measured but mirrored. Since the same-manufacturer comparison is the only evidence for site-specific variability beyond manufacturer or model differences, the symmetry assumption is load-bearing. The paper itself states in Section 5 that the assumption 'may not hold true due to the distinct configuration of the room and its structural characteristics.' No validation data from the left side are provided, and the y = 0 plane is interpolated between measured y = -5 and y = 25 cm planes. A small set of direct left-side measurements at the reported maximum-difference coordinates would decisively test whether the mirrored maps are trustworthy. Other concerns (unidentified scanner models, arbitrary fit functions, lack of raw data) are real but secondary; if the symmetry check passes, the core quantitative finding of inter-scanner differences remains plausible. Therefore the conditional verdict is appropriate; acceptance should require either symmetry validation or explicit removal of the same-manufacturer maxima from the headline claims.","tokens_in":12378,"tokens_out":10351,"duration_ms":96423,"concrete_test":"At each of the three sites, acquire direct measurements at ~10-20 points on the left side (x < 0) at and around the mirrored coordinates of the maximum-difference locations in Tables 3 and 4 (x = -70, z = 85, y = 0 and x = -50, z = -74, y = 0) using the same HP-01 gaussmeter and 10 cm grid protocol. Compare each measured left-side value to the mirrored right-side value from the published maps. If the deviation exceeds the reported uncertainty (0.15 T for the field maximum, 0.11 T/m for the gradient maximum) at any point, the symmetry assumption is falsified for that site and the same-manufacturer maximum differences are not empirically supported.","verdict_should_be":"UNCHANGED","load_bearing_attack":"Section 3 states that 'the intrinsic symmetries of the magnetic field were then leveraged' to mirror measurements from the right side into the left side, and Section 5 concedes that this 'may not hold true due to the distinct configuration of the room and its structural characteristics.' The same-manufacturer comparison (MRI-2 vs MRI-3) is the only evidence that site-specific factors matter beyond model or manufacturer differences, yet the maximum differences reported for this pair in Tables 3 and 4 lie on the unmeasured left side: the 0.17 T field maximum at x = -70, z = 85, y = 0 and the 0.47 T/m gradient maximum at x = -50, z = -74, y = 0. Moreover, y = 0 is an interpolated plane between the measured y = -5 and y = 25 cm planes. If the symmetry assumption fails, these headline same-manufacturer differences are products of the mirroring model, not direct measurements, and the paper's central safety implication ('even scanners from the same manufacturer can differ') loses its quantitative support.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper reports an experimental mapping of the static magnetic fringe field and its spatial gradient around three 3 T MRI scanners installed at three Italian hospitals, two of which are from the same manufacturer. Measurements were taken on a 10×10 cm grid on the right side of the patient table at three heights, then extended to full-room volumetric maps using parametric fitting, interpolation, and mirroring about the y-axis. The authors compare pairwise field and gradient differences and report maximum differences of up to 0.5 T and 1.64 T/m between scanners of different manufacturers, and up to 0.17 T and 0.5 T/m between scanners of the same manufacturer (Tables 3 and 4). They conclude that manufacturer-provided isogauss lines are insufficient for site-specific safety assessments.","tokens_in":12589,"tokens_out":4629,"duration_ms":49703,"significance":"If the quantitative claims were fully supported, the paper would be a valuable addition to MRI safety literature, providing concrete evidence that fringe fields vary across installations even for identical nominal field strength. The study's qualitative conclusion that scanners differ in their fringe fields is credible and consistent with prior work, and the direct measurement protocol near the gantry is a useful contribution. The principal limitation is that the headline quantitative maxima—especially the same-manufacturer differences—are not direct measurements but are generated by parametric fitting and by mirroring measured data across a symmetry plane that the authors themselves concede may not hold. The manuscript also introduces interpolation-uncertainty constants without calibration. These issues undercut the quantitative strength of the central claim, though they can be addressed in revision by restricting claims to measured regions, providing left-side measurements, or presenting the maxima as model-dependent estimates.","major_comments":[{"comment":"The maximum same-manufacturer differences in Tables 3 and 4 occur at negative x coordinates (e.g., Table 3, MRI-2 vs MRI-3, XZ: x=-70, y=0, z=85; Table 4, MRI-2 vs MRI-3, XZ: x=-50, y=0, z=-74). These positions lie on the left side of the scanner, which per Section 3 was not measured but generated by mirroring the right-side measurements under the assumption of symmetry about the y-axis. Section 5 concedes that this assumption 'may not hold true due to the distinct configuration of the room and its structural characteristics.' Consequently, the quantitative support for the claim that same-manufacturer scanners differ is not based on direct measurements at the reported maxima. Additionally, the y=0 plane used in the XZ rows of Tables 3 and 4 is interpolated between the measured y=-5 and y=25 cm planes. The authors should either measure the left side at the reported maxima, or restrict the quantitative comparisons to the directly measured side and clearly label mirrored regions as model-dependent.","section":"Section 3 (Data Processing) and Tables 3–4"},{"comment":"Equation (5) defines the interpolation uncertainty as σ_interp = σ_i + α × d_min with σ_i = 3% and α = 0.05%/cm, but these constants are asserted rather than derived or calibrated. They propagate into the total uncertainty in Eq. (6) and therefore into all error bars in Tables 3–6. Several reported maximum differences are comparable to or smaller than these uncertainties (e.g., Table 3, MRI-2 vs MRI-3, XZ: -0.17 ± 0.15 T; Table 4, MRI-2 vs MRI-3, YZ: 0.30 ± 0.17 T/m), so a different but equally plausible choice of σ_i or α could change which differences are statistically significant. The authors should justify the constants from data, for example through repeated measurements at varying distances, or provide a sensitivity analysis.","section":"Section 3, Eqs. (5)–(6)"},{"comment":"The field and gradient maps used to compute the maximum differences are not raw measurements but outputs of a two-stage procedure: each row and column is fitted with one of about ten parametric functions with five or six free parameters, with the model chosen by reduced χ² closest to unity, and then the outermost regions are filled by interpolation and mirroring. The manuscript does not report the spatial distribution of fit residuals, nor does it validate the fitted maps against held-out measurements. The largest same-manufacturer difference in Table 3 occurs at x=-70, z=85, which is in the extrapolated/mirrored region. The qualitative result that scanners differ is supported by direct measurements near the gantry, but the quantitative maxima should be presented with explicit caveats about model dependence, or supported by independent validation measurements.","section":"Section 3 (Model selection) and Section 4 (Tables 3–4)"}],"minor_comments":[{"comment":"Tables 3 and 4 appear to be mislabeled or displayed in reverse order: the data block with gradient values (labeled 'Table 4') appears before the data block with field values (labeled 'Table 3'), and both captions are printed consecutively before either table's data. Please reorder so that each table immediately follows its caption.","section":"Section 4, Tables 3 and 4"},{"comment":"Discussion item 3 refers to 'Tables 4 and 5' when the magnetic field and gradient differences are presented in Tables 3 and 4; please correct the cross-references.","section":"Section 5, Discussion item 3"},{"comment":"The text states that a minimum of 120 data points were collected per plane on a 10×10 cm grid, but a 10×10 grid over the red box in Figure 1 would contain far more than 120 points unless the grid covers only a partial region; please clarify the actual grid extent and point count.","section":"Section 2, Methodology"},{"comment":"Reference 3 shares the same DOI as Reference 2 (10.1007/s11517-021-02435-6); please verify and correct the DOI for the 2022 paper by Hartwig et al.","section":"References"},{"comment":"In Eq. (4), the symbol B_i is used without an explicit definition in the equation itself (it is defined only in the surrounding text); consider adding a notation explanation for clarity.","section":"Section 3, Eq. (4)"}],"recommendation":"major_revision","confidential_remarks":"The core idea is useful and the direct measurements near the gantry provide some support for the qualitative conclusion. However, the paper's most original quantitative claim—that same-manufacturer scanners differ by up to 0.17 T and 0.5 T/m—rests on mirrored, interpolated data at locations where the symmetry assumption is explicitly questionable. I would encourage the editor to request either left-side measurements or a substantial reframing of the claims as model-dependent estimates. The manuscript would also benefit from making the measurement data and scripts publicly available for the reviewers to verify the model selection and uncertainty constants."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Dear colleague,\n\nThe useful thing here is the dataset: fringe-field measurements at three 3 T sites, including two scanners from the same manufacturer, taken at three heights with a documented protocol. The paper makes the practical point that manufacturer isogauss lines are not enough for site-specific safety assessment, and the direct measurements near the gantry do show real differences.\n\nThe authors also deserve credit for reporting uncertainties, giving coordinates for the maximum differences, and acknowledging the main limitations up front. That is honest.\n\nThe soft spot is exactly the one the paper admits in Section 5: the full volumetric maps are completed by assuming left-right symmetry and mirroring measurements from the right side. The same-manufacturer comparisons, which are the most interesting part, have their headline maxima (0.17 T field, 0.5 T/m gradient) on that unmeasured mirrored side, and the y = 0 plane is interpolated between two measured planes. So those numbers are model outputs, not measurements. The paper states the symmetry assumption 'may not hold true' because of room structure, which is a fair concession, but the abstract still implies the comparison is purely measurement-based.\n\nThere is also a smaller overstatement: the abstract and introduction say 'identical MRI models,' but Table 1 only shows 'same manufacturer,' and no model names are given. That needs fixing.\n\nThe interpolation error constants in Eq. (5) are chosen rather than derived, so the reported error bars probably understate the true uncertainty in the mapped regions. That is a moderate concern, not a fatal one.\n\nNone of this undermines the central qualitative claim: the measured data on the right side alone show site-to-site differences that matter for worker exposure. But the specific maximum values for the same-manufacturer pair should be flagged as conditional on the symmetry and interpolation assumptions.\n\nThis paper is aimed at MRI safety professionals, and it gives them a concrete example of why generic isogauss lines are insufficient. It deserves a serious referee who can push for the raw data, a sensitivity check on the symmetry assumption, and a correction of the 'identical models' language. I would accept it for peer review with that in mind.\n\nBest,\n[Your name]","headline":"Useful dataset showing fringe-field variability across 3T sites, but the headline same-manufacturer differences rest on mirrored and interpolated data and need a clear caveat.","tokens_in":13138,"tokens_out":2611,"would_cite":false,"duration_ms":29946,"reading_group":"maybe","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":[],"pacs":[],"model":"deepseek-v4-flash","headline":"Measured fringe fields around three 3 T MRI scanners differ by up to 0.5 T between manufacturers and 0.17 T between same-brand units, so site-specific mapping, not vendor diagrams, is needed for safety assessment.","keywords":["MRI fringe field","3 Tesla scanner comparison","experimental magnetic field mapping","site-specific field distribution","MRI safety","occupational exposure","static magnetic field","spatial field gradient"],"falsifier":"Measure the fringe field on both the left and right sides of one of the three scanners on the same $10 \\times 10\\,\\mathrm{cm}$ grid at the same three heights; if field magnitudes at mirrored points differ by more than the combined uncertainty, the unmeasured half of the reported maps — and any maximum difference located there — is a model artifact rather than a measurement.","tokens_in":12190,"feed_emoji":"🧲","tokens_out":10505,"duration_ms":108041,"temperature":0.7,"pith_summary":"The paper tries to establish that the magnetic fringe field around a 3 Tesla MRI scanner is an installation-specific quantity, not a fixed property of the magnet or its nominal field strength. The authors measured the field on a 10 cm grid beside three clinical scanners, fitted and interpolated the data into three-dimensional maps, and computed spatial gradients from those maps. They found maximum differences of 0.5 T in field strength and 1.64 T/m in spatial gradient between scanners of different manufacturers, and still 0.17 T and 0.5 T/m between two scanners of the same manufacturer. If the result is correct, manufacturer-supplied isogauss line plans cannot define the zones where staff are exposed to high fields or steep gradients; each installation would need its own post-installation measurement.","feed_headline":"Measured 3 T MRI fringe fields vary by up to 0.5 T","feed_subtitle":"Even same-brand scanners differ by 0.17 T, so manufacturer safety maps are not enough.","key_machinery":"The argument is carried by a measurement-plus-interpolation pipeline rather than by a single equation. A commercial magnetometer records the field magnitude on a $10 \\times 10\\,\\mathrm{cm}$ grid on the right side of each scanner at three heights (waist, chest, and head/eyes); a nonlinear least-squares fitter selects among roughly ten parametric functions, including $f_{\\mathrm{exp}}(\\boldsymbol{b},x) = b_1 e^{-b_2 x} + b_3 e^{-b_4 x} + b_5 e^{-b_6 x}$ and variants with polynomial prefactors, using the reduced chi-squared $\\chi^2_{\\mathrm{red}}$ closest to unity. The fitted rows and columns are interpolated to volumetric maps, and the field is mirrored across the plane through the isocenter perpendicular to the patient table to double the number of available planes. The spatial gradient magnitude is then computed from partial derivatives of the interpolated field with a 1 cm step, and uncertainties are combined quadratically from instrumental, fitting, and interpolation errors. This pipeline is what turns sparse point samples into the comparative maps and into the reported maximum differences.","core_discovery":"On the paper's own terms, the discovery is that two 3 T scanners from the same manufacturer, installed in different hospitals, produce measurably different fringe fields and spatial gradients, and that the differences are larger still between manufacturers. The largest measured field difference was 0.5 T and the largest gradient difference was 1.64 T/m between different manufacturers; for the same-manufacturer pair the maxima were 0.17 T and 0.5 T/m. These maxima sit close to the gantry, in the zones where radiographers work during patient positioning. The authors conclude that the fringe field and spatial gradient differed across all sites despite the same nominal $B_0 = 3\\,\\mathrm{T}$, and that the generic information provided by manufacturers is insufficient for assessing worker exposure.","pith_inferences":["If the reflection symmetry across the plane perpendicular to the patient table is broken by room structure, the mirrored half of every map is a model artifact; a two-sided measurement would show whether the reported maxima are real, overestimated, or underestimated.","A practical extension the authors do not pursue is to couple these static gradient maps with recorded staff movement paths, because the biological exposure is set by the time-varying field along the worker's trajectory, not by the static map alone.","The three-site sample cannot separate shielding type from room geometry; repeating the protocol at sites with active versus passive shielding, and with documented structural layouts, would be needed to isolate the cause of the inter-manufacturer differences.","Because the maps were built from only three heights on one side, vertical gradients between planes are less constrained than horizontal ones; denser height sampling would tighten the volumetric reconstruction and test the symmetry assumption directly."],"forward_implications":["Safety zones drawn from manufacturer isogauss projections can be wrongly sized or placed, so commissioning of a 3 T scanner should include post-installation field mapping.","Two scanners of the same model and manufacturer can expose staff to measurably different fields, meaning room design and shielding, not nominal field strength, set the real fringe field.","The largest observed discrepancies sit at the gantry edge, exactly where technologists position patients, so the worst-case exposure point is where generic diagrams are least informative.","The measurement and interpolation protocol generalizes to other MRI suites once a minimum $70 \\times 70\\,\\mathrm{cm}$ measurement area is covered, allowing comparable maps and gradients to be produced elsewhere."],"supporting_citations":[{"why":"Supplies the validated multi-plane measurement protocol and analysis approach that the study applies at all three sites.","marker":"Hartwig et al. (9)"},{"why":"Prior characterization work showing that manufacturer isogauss lines under-resolve the high-gradient zone near the gantry, motivating on-site measurement.","marker":"[8], [9]"},{"why":"Establish that moving through the fringe-field gradient creates a time-varying field exposure with physiological effects, making gradient maps safety-relevant.","marker":"[2], [3], [4]"},{"why":"Documents that vendor field projections assume no supplementary shielding, supporting the claim that real installations can differ.","marker":"[10]"},{"why":"Provides the standardized-residuals method used to assign uncertainty to fitted field points.","marker":"[11]"},{"why":"Provides the uncertainty-propagation formula used for interpolation error in the volumetric maps.","marker":"[12]"},{"why":"Support the interpretation that inter-manufacturer and inter-site fringe-field differences arise from shielding configuration.","marker":"[14], [15]"}],"fun_headline_variants":["3T MRI fringe fields vary up to 0.5 T across scanners, study finds","Measured: 3T MRI fringe fields differ by up to 0.5 T across sites","Even same-brand 3T MRI scanners differ by 0.17 T in fringe fields","Manufacturer safety maps insufficient: 3T MRI fringe fields differ by 0.5 T","Site-specific shielding causes 3T MRI fringe field gaps up to 0.5 T"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The load-bearing premise is that the magnetic field is mirror-symmetric across the plane through the isocenter perpendicular to the patient table, so the left half of every map is not measured but copied from the right; the authors acknowledge that room layout and structural features may break this symmetry, and if it breaks, every field and gradient value on the unmeasured side is a model artifact.","fun_headline_variants_meta":{"raw":{"variants":["3T MRI fringe fields vary up to 0.5 T across scanners, study finds","Measured: 3T MRI fringe fields differ by up to 0.5 T across sites","Even same-brand 3T MRI scanners differ by 0.17 T in fringe fields","Manufacturer safety maps insufficient: 3T MRI fringe fields differ by 0.5 T","Site-specific shielding causes 3T MRI fringe field gaps up to 0.5 T"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.001009,"raw_usage":{"total_tokens":4211,"prompt_tokens":836,"completion_tokens":3375,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":452,"completion_tokens_details":{"reasoning_tokens":3256}},"tokens_in":452,"tokens_out":3375,"duration_ms":25810,"temperature":1.0,"reasoning_tokens":3256,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-06T13:13:51.406827+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Measure the fringe field on both the left and right sides of one of the three scanners on the same $10 \\times 10\\,\\mathrm{cm}$ grid at the same three heights; if field magnitudes at mirrored points differ by more than the combined uncertainty, the unmeasured half of the reported maps — and any maximum difference located there — is a model artifact rather than a measurement.","supporting_citations":[{"cited_title":"Health complaints among nurses working near MRI scanners — A descriptive pilot study","cited_arxiv_id":null,"evidence_quote":"Documents that vendor field projections assume no supplementary shielding, supporting the claim that real installations can differ."},{"cited_title":"The Procedure for Quantitative Characterization and Analysis of Magnetic Fields in Magnetic Resonance Sites for Protection of Workers: A Pilot Study","cited_arxiv_id":null,"evidence_quote":"Provides the standardized-residuals method used to assign uncertainty to fitted field points."},{"cited_title":"Assessment of Exposure to Spatially Varying Magnetic Fields in MRI Environments: Modeling Analysis for Simulation Tools","cited_arxiv_id":null,"evidence_quote":"Provides the uncertainty-propagation formula used for interpolation error in the volumetric maps."}],"review_version":1}